Conceptualizing Representational Semantics: A Multiple Layered Spatial Data Integration Framework Based on Ontology

Abstract

Domain ontologies have been used as an effective mean to reconcile the heterogeneities between different spatial data sources. Most domain ontologies focus on the concepts from real world, but do not specify the semantics of the way to represent them. However, heterogeneities exist in forms of different frameworks(standards), data models (schemas), measurement backgrounds, algorithms, spatial-temporal features, etc, which makes seamless data integration a non-trivial work. Even if we have solved the problem of identifying domain concepts from different sources, we still have to deal with fusing the data associated with the same concept but from different representations. Representational semantics are about the concepts which give the semantic of how we measure and organize the result of observation of concepts from real world, e.g. property and relation, unit, coordinate system and process model. They are, while always implicit in all kinds of data and systems, very important for engineers responsible for the data management and integration, and users who want to evaluate the viability of the data for a specific task, for their indispensability for supporting semantic alignment, data structure conversions, and mathematical computations, all of which are crucial steps toward a successful integration of spatial databases. In this paper we argued that both domain ontologies and their conceptual representations are essential to spatial data integration and assuming complimentary roles. We proposed to capture and formalize representational semantics for the spatial data, together with domain ontologies, to facilitate semantic-dependent data integration with heterogeneous data sources and processing models. By separating the two-level ontologies, more restrictions and axioms can be added to them, the enhanced reasoning ability can improve the seamless spatial data integration. We studied several cases to show how to leverage representational ontologies to solve the data integration problems.